Postmortem · 2026-04-24

BB Squeeze SHORT — why our 2-year backtest was wrong by 59.6% in 2 months

Honesty-first report. We ran one strategy live on OKX for 55 days. The backtest said +49.9%; the live account returned -9.7%. Here's exactly what happened, what we changed, and what the current recommended list is.

Backtest (2 yr)

+49.9%

Live (55 days)

-9.7%

Gap

59.6%

The raw numbers

  • Strategy: BB Squeeze SHORT
  • Live period: 2026-01-13 → 2026-03-09 (55 days, 1,914 trades)
  • Live win rate: 52.7% · Profit factor: 0.88 · Max drawdown: 16.5%
  • 2yr backtest at same SL 10% / TP 8% / Top 10: PF 1.17 · +83% return
  • Live capital: $3,102 start → $2,800 end

What went wrong

BB Squeeze SHORT enters short after a Bollinger Band contraction + bearish breakout with volume confirmation. The 2-year backtest (2023–2025) had plenty of sustained down-trends where mean-reversion was rare — a good regime for a breakout-short.

2026-01 to 03 was different. Spot was choppy with frequent false-breakdown-then-reclaim. The strategy entered short on breakouts that reversed within 2–4 candles; TP never hit, SL frequently did. Win rate stayed near 53% (close to backtest), but the loser-to-winner magnitude ratio inverted — average loss (10%) exceeded average win (7%), dragging PF below 1.

In one sentence: the backtest underweighted the 2026-Q1 regime where volatility compressed without directional follow-through.

What we changed

  • Removed BB Squeeze SHORT from the recommended Quick Start. Its 2yr backtest is still strong (PF 1.17), but with no currently-winning live data we don't promote it.
  • Added walk-forward as a gate. Future promotions require at least one out-of-sample window passing, not just a flat 2yr total.
  • Paused single-strategy live tracking. While auto-trading is being hardened, we're not running any one strategy live — so no gap chart on the homepage would be misleading in either direction.
  • Kept the failure public. This page stays up. If we delete it, the pattern repeats.

What we recommend now

The Quick Start presets on /simulate show 5 backtest-verified strategies with honest metrics measured against the live backtest engine:

  • ATR Breakout (short) · PF 1.31 · Sharpe 0.98 · MDD 45.6%
  • Ichimoku Bearish (short) · PF 1.21 · Sharpe 0.85 · MDD 42.2%
  • BB Squeeze SHORT · PF 1.17 · (2yr backtest; live paused)
  • Keltner Fade (short) · PF 1.14 · Sharpe 0.71 · MDD 44.8%
  • MA Cross (both) · PF 1.09 · Sharpe 0.54 · MDD 33.5% (lowest)

Every number is reproducible — clicking the card runs the backtest through the exact same engine and returns the same metrics (±2% from cache refresh).

Try the verified list

ATR Breakout is the strongest (PF 1.31) but MA Cross is the lowest-drawdown (33%) if you're protecting capital. Both click-to-run.

Try ATR Breakout → Try MA Cross → How we verify

Published 2026-04-24 · Updated as more live data accumulates